[SciPy-user] problem with fmin_cobyla

Matthieu Brucher matthieu.brucher@gmail....
Fri Sep 7 07:22:18 CDT 2007


>
> The second example, as I said, does not violate any
> constraints, but is not the optimal solution, tho it
> seems that perhaps fmin doesnt look for the most
> minimal solution it can find? interestingly, I can
> replicate the second result even with more function
> calls, closer initial guesses, larger or smaller rho
> params, etc... so seems like it should be a real
> minima.
>

The usual optimization procedures are only local optimizers, they cannot
find the global optimizer, only a local one. Global optimization is a very
hard field as finding the correct set of parameters requires most of the
time an exhaustive search.

Matthieu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://projects.scipy.org/pipermail/scipy-user/attachments/20070907/fb658ee3/attachment.html 


More information about the SciPy-user mailing list